deepset-ai/haystack · error

query_embedding should be a non-empty list of floats.

Error message

query_embedding should be a non-empty list of floats.

What it means

InMemoryDocumentStore.embedding_retrieval raises ValueError when query_embedding is an empty list or its first element is not a Python float. The check `not isinstance(query_embedding[0], float)` fails fast before any similarity computation.

Source

Thrown at haystack/document_stores/in_memory/document_store.py:816

        top_k: int = 10,
        scale_score: bool = False,
        return_embedding: bool | None = False,
    ) -> list[Document]:
        """
        Retrieves documents that are most similar to the query embedding using a vector similarity metric.

        :param query_embedding: Embedding of the query.
        :param filters: A dictionary with filters to narrow down the search space.
        :param top_k: The number of top documents to retrieve. Default is 10.
        :param scale_score: Whether to scale the scores of the retrieved Documents. Default is False.
        :param return_embedding: Whether to return the embedding of the retrieved Documents.
            If not provided, the value of the `return_embedding` parameter set at component
            initialization will be used. Default is False.
        :returns: A list of the top_k documents most relevant to the query.
        :raises ValueError: if filters have invalid syntax.
        """
        if len(query_embedding) == 0 or not isinstance(query_embedding[0], float):
            raise ValueError("query_embedding should be a non-empty list of floats.")

        if filters:
            InMemoryDocumentStore._validate_filters(filters)
            all_documents = [
                doc
                for doc in self.storage.values()
                if document_matches_filter(
                    filters=filters, document=doc, strict_datetime_comparison=self.strict_datetime_comparison
                )
            ]
        else:
            all_documents = list(self.storage.values())

        documents_with_embeddings = [doc for doc in all_documents if doc.embedding is not None]
        if len(documents_with_embeddings) == 0:
            logger.warning(
                "No Documents found with embeddings. Returning empty list. "
                "To generate embeddings, use a DocumentEmbedder."

View on GitHub (pinned to e318778c9b)

Solutions

1. Call .tolist() on numpy arrays: store.embedding_retrieval(query_embedding=model_output.tolist())
2. Ensure the embedder returns Python floats and that the input text is non-empty
3. Cast explicitly: query_embedding=[float(x) for x in raw_embedding]
4. If using quantized/int embeddings, keep them as Python floats or convert per the store's expectations

Example fix

// before
embedding = model.encode(text)  # numpy array of np.float32
results = store.embedding_retrieval(query_embedding=embedding)
// after
results = store.embedding_retrieval(query_embedding=embedding.tolist())

def is_valid_query_embedding(e) -> bool:
    return bool(e) and isinstance(e, list) and all(isinstance(x, float) for x in e)

if not is_valid_query_embedding(query_embedding):
    query_embedding = [float(x) for x in query_embedding]  # normalize (also converts numpy floats)

def is_float_list(x) -> bool:
    return isinstance(x, list) and len(x) > 0 and isinstance(x[0], float)

try:
    docs = store.embedding_retrieval(query_embedding=query_embedding)
except ValueError as e:
    if "query_embedding should be a non-empty list of floats" in str(e):
        docs = store.embedding_retrieval(query_embedding=[float(x) for x in query_embedding])
    else:
        raise

When it happens

Trigger: Passing query_embedding=[]; passing a list of ints (e.g. quantized embeddings); passing a list of numpy floats such as np.float32/np.float64 scalars from an embedder that returns numpy output, since np.floating is not a subclass of Python float.

Common situations: Embedder returning an empty list for empty input text, int8/binary quantized embeddings, forgetting to call .tolist() on a numpy array before passing it in.

Related errors

• Query should be a non-empty string
• The embedding size of the query should be the same as the em
• Hook of type '{type(h).__name__}' is registered under hook p
• Tool execution requires at least one tool.
• The following tool names are not valid: {invalid_tool_names}

AI-assisted analysis of deepset-ai/haystack@e318778c9b (2026-08-30). Data as JSON: /api/errors/156ea530e6a55060. Report an issue: GitHub.